Home / Browse / Steel Bank Common Lisp / Mail / Archive

Steel Bank Common Lisp

Email Archive: sbcl-help (read-only)

[Sbcl-help] gc bug

From: Faheem Mitha <faheem@fa...> - 2013-01-01 09:26

Hi everyone and happy new year,

About 6 months I posted a bug report: 
https://bugs.launchpad.net/sbcl/+bug/1009267 ("(gc :full t) has no effect 
under certain circumstances"). This bug has attracted little interest, 
despite being, in my opinion, quite significant. The bug was marked 
incomplete. I added more details as requested, but it is still marked 
incomplete. I have not even got an official acknowledgement that this bug 
is reproducible. I am at this point quite confident this is real, since I 
have been able to reproduce it on other machines, besides my home machine, 
different versions of SBCL, and also a friend who does not know CL was 
able to reproduce it following my instructions. I just checked, and this 
is still reproducible for me on 1.1.2. As I say in the report, the most 
significant thing about this bug is that it breaks (gc :full t).

See also https://bugs.launchpad.net/sbcl/+bug/936304. As I comment there, 
that (similar) bug is also still reproducible, even though it is marked as 
closed.

This bug only shows up under relatively special cirumstances. However, 
these relatively special circumstances (large float arrays being created 
and destroyed) are not uncommon in scientific computing. It is also worth 
emphasizing that I can only reproduce this bug on x86. I'm not sure why it 
does not show up on amd64, but that being so, it may reduce the number of 
people who are exposed to this bug.

I realise that this bug may be considered low priority, and also may be 
difficult to fix. However, I would appreciate at least a confirmation that 
it is reproducible.

                                                      Regards, Faheem Mitha

--
Under the two moons of Hydrot, and under the eternal stars, the
two-inch wooden spaceship and its microscopic cargo toiled down the
slope toward the drying little rivulet.
--
James Blish - Surface Tension

Thread View

Thread	Author	Date
[Sbcl-help] gc bug	Faheem Mitha <faheem@fa...>

Re: [Sbcl-help] gc bug

From: Christophe Rhodes <csr21@ca...> - 2013-01-01 10:35

Faheem Mitha <faheem@...> writes:

> I realise that this bug may be considered low priority, and also may be 
> difficult to fix. However, I would appreciate at least a confirmation that 
> it is reproducible.

For what it's worth, I just reproduced the symptoms lp#1009267.
(Equally for what it's worth, I won't have any time to look at
understanding what the problem is, let alone fixing it, in the
forseeable future.)

Best wishes,

Christophe

Re: [Sbcl-help] gc bug

From: Faheem Mitha <faheem@fa...> - 2013-01-01 11:37

On Tue, 1 Jan 2013, Christophe Rhodes wrote:

> Faheem Mitha <faheem@...> writes:

>> I realise that this bug may be considered low priority, and also may be 
>> difficult to fix. However, I would appreciate at least a confirmation 
>> that it is reproducible.

> For what it's worth, I just reproduced the symptoms lp#1009267.
> (Equally for what it's worth, I won't have any time to look at
> understanding what the problem is, let alone fixing it, in the
> forseeable future.)
>
> Best wishes,
>
> Christophe

Ok. Thanks, Christophe.

        Regards, Faheem

Re: [Sbcl-help] gc bug

From: Faheem Mitha <faheem@fa...> - 2013-01-01 12:18

On Tue, 1 Jan 2013, Christophe Rhodes wrote:

> Faheem Mitha <faheem@...> writes:
>
>> I realise that this bug may be considered low priority, and also may be
>> difficult to fix. However, I would appreciate at least a confirmation that
>> it is reproducible.
>
> For what it's worth, I just reproduced the symptoms lp#1009267.
> (Equally for what it's worth, I won't have any time to look at
> understanding what the problem is, let alone fixing it, in the
> forseeable future.)

Hi Christophe,

If it is possible for you to update the bug report status, that would be 
helpful. Currently, this bug report is still listed as "Incomplete".

                                                          Regards, Faheem

Re: [Sbcl-help] gc bug

From: Christophe Rhodes <csr21@ca...> - 2013-01-01 12:27

Faheem Mitha <faheem@...> writes:

> If it is possible for you to update the bug report status, that would
> be helpful. Currently, this bug report is still listed as
> "Incomplete".

OK, done.